Pathology AI scribe 2026: surgical specimens, cytology, synoptic reports, molecular, and the CAP standards reality

May 8, 2026 · 6 min read

Pathology generates the most structured documentation in medicine. CAP (College of American Pathologists) synoptic reports are required for cancer specimens and have specific element checklists per organ + cancer type. Cytology has its own structured framework (Bethesda for Pap, Milan for salivary, Paris for urinary). Molecular profiling reports must integrate with downstream oncology decisions. AI scribes built around clinical visit narratives don't fit pathology's reporting workflow at all — but voice-driven structured report generation is highly relevant.

The 2026 pathology-aware AI scribe stack handles four things general scribes miss: synoptic report generation per CAP cancer protocols, cytology structured reporting per system-specific frameworks, molecular profiling integration, and gross / microscopic dictation with structured output.

CAP synoptic cancer protocols

CAP publishes synoptic protocols for each cancer type with required data elements. Examples:

• Breast carcinoma: tumor size, histologic type, grade, hormone receptors, HER2, margins, nodes
• Colon adenocarcinoma: tumor location, size, histologic grade, lymphovascular invasion, perineural, MMR/MSI, pT/pN/M
• Prostate adenocarcinoma: tumor extent, Gleason score (primary + secondary + tertiary if applicable), perineural, EPE, margins
• Lung NSCLC: histologic type (subtype), grade, pleural invasion, lymphovascular, margins, pT/pN/M, biomarkers (PD-L1, EGFR, ALK, ROS1, etc.)

An AI scribe that helps the pathologist dictate by surfacing the next required CAP element accelerates report generation and reduces missed elements.

The pathology system prompt

You are documenting a pathology report from dictation.

INPUT:
- Pathologist dictation transcript
- Specimen accession data: organ, procedure type, clinical history
- For cancer specimens: applicable CAP synoptic protocol

OUTPUT structured report:

For surgical pathology cancer:
1. Specimen identification + procedure
2. Clinical history
3. Gross description (specimen received, fixation, dimensions, cut surface, lesion description)
4. Microscopic / final diagnosis
5. CAP synoptic data elements (per cancer-type protocol):
   - Tumor type + grade
   - Tumor size + extent
   - Margins (named with distance)
   - Lymphovascular / perineural invasion
   - Lymph nodes (involved / examined, ECE)
   - Stage (pT/pN/M with edition number)
   - Cancer-type-specific biomarkers
6. Comments / clinical correlation
7. Sign-out date and pathologist

For cytology:
1. Specimen + procedure
2. Adequacy assessment
3. Diagnostic interpretation per applicable framework (Bethesda/Milan/Paris/IAC)
4. Comments
5. Recommended follow-up if applicable

For molecular profiling:
1. Specimen + tumor tissue percentage
2. Methodology (IHC, FISH, NGS panel)
3. Genes/markers tested
4. Findings with clinical interpretation
5. Treatment-actionable mutations highlighted

For non-cancer surgical:
- Standard gross + microscopic + diagnosis structure

Cite transcript or specimen data. For CAP synoptic, ensure all required elements are populated — flag missing elements before sign-out.

Voice-driven dictation workflow

Pathologists typically dictate gross descriptions at the cutting bench and microscopic findings at the scope. The AI scribe captures voice continuously and structures into the report skeleton in real time. Clinician reviews and signs.

Vendor and DIY paths

For pathology labs, native LIS (laboratory information system) platforms have synoptic templates. The DIY stack with custom CAP protocol prompt outperforms general voice-to-text on report quality and completeness. Solo pathologists or small group practices benefit most.

BAA chain

Lab + LIS + voice capture + transcription + LLM vendor.

When to start

For pathology practices generating > 50 cancer cases per week, structured synoptic reporting reduces missed elements and accelerates sign-out. The DIY stack with CAP protocol embedding is buildable in 4-6 weeks.